Here's a long-form exercise:

1. Break up into small groups (2-5 people)
2. Someone in each group, picked at random, presents an upcoming problem or decision.
   The other participant(s) ask questions to clarify the problem/decision.
   The problem/decision can be either real or imaginary, but if imaginary the presenter must come up with appropriately detailed answers to questions.
3. Everyone collaborates in generating a list of possible solutions. Five is plenty.
4. Everyone privately notes their preferred solution.
5. Everyone collaborates on a list of expected consequences of each solution (with confidence intervals, if desired) and publicly announce which consequence-list they consider preferable.
6. Everyone reveals their earlier preference from step 4, and if it's different from their preference from step 5 explains what changed.

This will probably take a fair amount of time.
The shorter-form version starts with steps 1-3 already done as example problems.
It probably helps if step 6 is a surprise.

Incidentally, I prefer "Compare Likely Consequences" to "Check Consequentialism" as a label for the skill.